A computer system that includes a plurality of processing elements for parallel computation utilizes a free-space optical network for communication between the processing elements. Such a network employs an optoelectronic switch that includes a binary H-type tree for routing signals to selected ones of an array of lasers. Hybrid repeaters are included in the optical paths to generate output optical beams colinear that are with incident input beams.
|
1. In an information processing system in which a plurality of processing elements that are to be interconnected are grouped in a plurality of clusters, and in which communication between clusters is provided by light beams passing along optical channels formed between communicating clusters, a light regenerator for diverting light beams in selected optical channels for regeneration and return to said optical channels comprising:
light reflecting means comprising a pair of isoceles right angle prisms of a material transparent to said light beams coupled together at an interface along which are provided at least one reflecting facet for selectively deflecting transversely out of their optical channels light beams to be regenerated; means in the path of said reflected light beams for detecting said light beams for deriving electrical signals representative of the data on said light beams, and amplification of said electrical signals, light emitting means supplied with said derived electrical signals for producing light beams that are amplified replicas of said selected light beams whose axes are parallel to but displaced from said deflected light beams, and means for displacing the axes of said regenerated light beams and directing the beams towards said first-mentioned light reflecting means for reflecting each of said beams back into optical channels colinear with their original optical channels.
2. A light regenerator in accordance with
3. A light regenerator in accordance with
4. A light regenerator in accordance with
|
|||||||||||||||||||||||||||
This is a division of application Ser. No. 08/340,680 filed Nov. 16, 1994, now U.S. Pat. No. 5,619,359.
This invention relates to optoelectronic apparatus and more particularly to such apparatus for use in information handling systems.
As information handling systems become more complex, and require more information to be handled at greater speeds, they need components of more compact size, greater capacity, and faster speeds. It is widely believed that optoelectronic components are particularly promising for use in such systems because optics offers considerable advantages over pure electronics as a communication medium, as data rates move into the gigabit range. For example, it now appears that free space optics offers considerable advantages over wire leads for the interconnection of processing elements in computer systems that use closely spaced semiconductive chips as the processing elements, as in computers that involve massively parallel computation.
Such optical interconnection networks depend importantly on optical components that are compact, rugged, efficient and sealable for use in large systems. The present invention is directed at providing such apparatus.
The present invention is its preferred embodiments focuses particularly on devices adapted to function in a parallel computation system of the kind described in our copending application Ser. No. 07/994,220, filed Dec. 18, 1992 having the same assignee as this application. However, such devices should also be useful in various other forms of high speed information handling systems.
In that application, there is described a parallel processing system that divides the processing elements into clusters and includes an interconnect network that is of a three-stage modified-clos type. It uses electronic switches in the first and third stages for switching between processing elements in a common cluster and optoelectronic switches to switch between clusters in the second stage that use free-space light beams for use as the transmission medium between clusters. This system is designed so that generally only two routing decisions need to be made in interconnecting any two processing elements. The present invention provides components that are particularly well adapted for use in the second stage of such a system. Reference can be made to such copending application for a more detailed description.
In such a system, the second stage comprises a plurality of lasers that serve as transmitters and a plurality of photodetectors that serve as receivers, the communications therebetween being by light beams that originate from the lasers and terminate at the photodetectors, the routing therebetween being done by various optical elements.
For such communication, an optoelectronic switch is advantageously employed by means of which one laser is an array of lasers is selectively energized by a transmitting processing element, the selection being dependent on the cluster to which belongs the photodetector of the processing element intended as the destination of the information being transmitted.
In one aspect, the present invention is a device that serves as a suitable optoelectronic switch that permits such selection.
Advantageously, each processing element includes both an information processor and a routing portion and all the routing portions in a cluster are interconnected by way of a local switch that permits each processing element to communicate with all other processing elements in a cluster.
Additionally, information that is generated in the information processor of a processing element and intended for transfer to a different cluster may be transferred by way of its routing portion to the routing portion of a different processing element in the same cluster for transfer by the optical interconnection network.
As a consequence, each routing portion has available to it, for transfer by the optical interconnection network, information from the information processor of each of the processing elements in the cluster.
By use of the routing portions, it becomes possible to couple the information from any processing element of the cluster to any optoelectronic switch in the same cluster.
In a similar fashion, advantageously each photodetector in a cluster is coupled to a routing portion of a processing element, and so can couple to the routing portion of any processing element in the cluster, whereby the information received by the photodetector in the cluster can be transferred to any processing element in the cluster for use by it.
Moreover, a light beam, in its routing from the laser associated with the cluster of the transmitting processing element to the photodetector associated with the cluster of the receiving processing element, may experience considerable attenuation since such routing may involve passage through a number of microlenses and reflection from a number of partially transmissive mirrors.
Accordingly, in another aspect, the present invention is a regenerator that can be inserted in the path of a weak light beam and then used to intercept the light beam, amplify or regenerate it, and send it along on a path that is collinear with the path from which the beam was initially intercepted.
As mentioned above, in one aspect the invention is a novel optoelectronic switch. This switch is designed to be supplied by an input signal intended for a particular destination, which destination is used to determine which laser of an array of lasers is to be intended by the input signal. In a particular embodiment, the lasers are arranged in a two-dimensional array in a first semiconductive chip and the input signal is applied to a second semiconductive chip in which there has been formed a switch, advantageously of a binary H-tree configuration, whose leaves form a two-dimensional pattern that matches the two-dimensional pattern of the laser anodes of the first chip. Accordingly, the first chip can be butted against the second chip and there bonded to form electrical contacts between the pads at the ends of the leaves of the H-tree switch in the second chip and the anode electrodes of the lasers of the first chip. The address information associated with the input signal is used by the switch to select the laser appropriate to the intended destination of the signal information. Both chips typically might be of gallium arsenide. However, it is anticipated that soon it will be feasible to integrate the two chips into a monolithic chip. Such an optoelectronic chip should have application in various systems other than that described in the copending application. Additionally, each laser may be of the same wavelength or many different wavelengths may be involved.
In accordance with another aspect of the invention, a light regenerator for insertion in the path of the light beams is formed by a hybrid detector/transmitter that is provided with correction for the detector/transmitter physical position offset. Such correction may be provided in a variety of ways including either the use of v-groove mirrors to compensate for the original detector/laser offset or the addition of a second detector/transmitter that provides a second compensating offset, or the use of controlled polarizations and a birefringent element to provide a compensating offset.
In another aspect, the invention is an arrangement for coupling light beams of different wavelengths into and out of optical channels for use in a wavelength division multiplexing arrangement.
The invention will be better understood from the following more detailed description taken in conjunction with the accompanying drawing in which:
FIG. 1 shows schematically the basic elements of a free space optical interconnection network of the kind to which the invention is primarily applicable.
FIG. 2 is a three dimensional exploded view of an optoelectronic switch in accordance with one aspect of the invention.
FIGS. 3, 4 and 5 are cross sectional view of three different embodiments of regenerators in accordance with the second aspect of the invention.
FIG. 6 shows an arrangement for cooling a plurality of beams of different wavelengths into and out of free space optical channels in accordance with another aspect of the invention.
With reference now to the drawing, the parallel computation system 10 contains a number of columns 11, each column including a plurality of circuit boards 12, each circuit board including a cluster of processing elements 13. The processing elements are advantageously divided among the various boards such that during computations the processing elements between which the communication is expected to be appreciable are located as a cluster on a common board and the communication between processing elements on a common board is controlled by electronic circuitry on the common board. However, communication between clusters on different circuit boards is done optically to avoid the complexity of providing wire leads between the boards, which could be complex when hundreds of circuit boards, each with tens of processing elements are involved, and the destination of signals between processing elements might need to be changed rapidly.
To provide optical communication between the boards each circuit board is provided with a number of optoelectronic switches 15. In accordance with the invention, each optoelectronic switch in a monochromatic system is designed to include a two-dimensional array of lasers in which, each emits light of the same wavelength, but each is positioned to route its beam to a different photodetector associated with a distant circuit boards. The output of a processing element which is to be directed to the input of a processing element not on the common circuit board is applied to the input of the optoelectronic device and switched to the laser aimed at a distant photodetector on the appropriate circuit board for accessing the desired processing element. Then at the receiving end, the output of the receiving photodetector can be routed electronically to the desired processing element. To this end, each circuit board is associated with both a plurality of optoelectronic switches 15 to serve its transmitting processing elements and an array of photodetectors 16 to serve its receiving processing elements. Appropriate deflecting elements 17 are used to deflect the beams emitted by the optoelectronic switches into appropriate paths for travel to their intended destinations and to deflect the beams from their travel paths into the appropriate photodetectors. Additional deflecting elements 18 are located at the ends of each column for routing the beams appropriately through the various columns. Microlenses (not shown) are also typically included along the beam paths. Also included at various points along the beam paths are repeaters 40 to be described more fully below.
In FIG. 2 there is shown a three-dimensional exploded view of an optoelectronic switch 15 in accordance with the invention for use at the transmitting end as the source of data streams for routing in a particular channel.
As seen, the switch 15 includes two principal components, a chip 20 which houses the electronic circuitry that provides the routing of the input data to a particular diode laser and a chip 30 which includes an array of lasers of which one is to be selected for emitting a beam of light aimed at a photodetector associated with the cluster of the processing element with which the transmitting processing element is to communicate.
To this end, the 20 chip houses a binary tree switch, preferably of the known H-type configuration, whose input terminal 21, corresponding to the root of the tree is supplied with the signal information to be transmitted by the sending processing element. The tree includes various branches 23 arranged in a common plane so that no branch crosses over another so that pads 26 at the leaf ends form a two-dimensional array, advantageously a square array, as shown. Each of the leaf pads 26 is provided with a solder bump that serves as a connection pad to the tree. The chip 20 further includes a plurality of individual switches 27 that can be set to a high or low resistance state by control signals. The switches can be field-effect transistors connected between branch points of the tree. The switches are appropriately set by the address of the intended destination to route the signal information to the desired leaf of the tree. The signal information is then used to modulate the light beam emitted by the laser to which the chosen leaf is coupled. The chip 20 may also house the driver circuits for proper bias and modulation of the emitting laser.
The second component 30 of the switch 10 comprises a chip in which there has been formed a two-dimensional array of diode injection lasers that are surface emitting to emit a beam substantially circular in cross-section to simplify the optics for the routing of the beams. The anode electrodes 31 of the lasers form a two-dimensional square pattern that closely matches the two-dimensional square pattern of the solder bumps at leaf pads 26 so that when chip 20 is bonded to chip 30 low-loss electrical connections are formed between the binary tree leaf pads and the laser array anode electrodes. A common cathode can be used for all the lasers in the chip.
At some future date, it may become feasible to make the switch 15 essentially all in a monolithic chip, such as a chip essentially of gallium arsenide or other suitable semiconductor. In such a case by appropriate positioning of the various elements in the chip, there may be reduced the need for electrode connections in the chip. Of course, for some applications, it may be feasible to use other materials, such as silicon, for chip 20.
With reference back to FIG. 1, it will be important to include various optical elements, such as microlenses for optically relaying the various laser beams and deflecting elements for routing the various beams between the lasers and the photodetectors. It can be appreciated that some of these paths may be relatively long during which their beams can suffer considerable attenuation.
To compensate for such possible attenuation, it is important to amplify or regenerate some of the beams. It is also important, if the optics is to avoid becoming too complex, to provide such amplification or regeneration in a manner that permits the regenerated beam to maintain essentially the same path it had before amplification or regeneration.
Unfortunately at the present state of technology, there is not available a practical form of colinear optical amplifier so that it is necessary to use hybrid forms of regenerators, usually known as repeaters, that involve optic-to-electronic conversion followed by electronic-to-optic reconversion. Because such operations consume power, it is preferably to regenerate selectively only the weakest signals in a given repeater array and to include multiple repeater arrays in the network.
In FIG. 1, there are also shown included repeaters 40 at various points, advantageously either at the ends of selected columns, or in the paths between columns. The repeaters may take various forms, each of which however is designed to produce an output beam which is not only an amplified version of the input beam but also is colinear with the input beam.
One embodiment of such a repeater is shown in FIG. 3. The repeater 50 includes a mirror assembly comprising the pair of right angle isoceles prisms 51, 52 positioned to form a plane interface 53 which is provided with at least one localized reflecting facet 54 where the weak beam A to be regenerated would otherwise pass through the interface. As shown, at the facet this beam is reflected vertically to be incident on a photoreceiver 55 that converts the beam into an electrical signal that is amplified and then used to modulate a laser 56. This laser is positioned to emit a beam that is displaced horizontally from the reflecting facet 54 and that passes largely unaffected through the mirror assembly to be incident on a second photoreceiver 57 that converts the incident beam into an electrical signal. This signal in turn is used to excite laser 58 that generates a new beam designed to pass partially through the mirror assembly but is redirected as it reaches the interface 53 the localized reflecting facet 54 that redirects the incident beam along a path that forms a continuation of the path the beam would have taken through the mirror assembly had it not suffered the initial redirection. As a consequence, the beam now regenerated passes along a collinear extension of its original path. Moreover, a beam not needing regeneration, such as beam B, passes through the interface 53 at a point free of any reflecting facet and so is little affected by passage through the mirror assembly.
It is, of course, possible to include several reflecting facets at interface 53 to deflect additional beams for regeneration in similar fashion to form a repeater array for a number of beams.
An alternative assembly is shown in FIG. 4. The repeater 50 differs from that of FIG. 3 in that in place of the second photoreceiver-laser pair, it employs a polarizing micromirror facet 61 at the interface 62 of mirror assembly 63, a birefringent element, such as a calcite plate 64, a quarter wave plate 65 and mirror 66, for displacing the regenerated beam A provided by the photoreceiver-laser pair 67 sufficiently that it is redirected along its original path.
Similarly, as still another embodiment, that may prove to be preferred FIG. 5 shows a repeater 70 that utilizes a V-groove reflecting mirror 71 to displace the regenerated beam A provided by the photoreceiver-laser pair 72 sufficiently that it is redirected along it original path after reflection in reflecting facet 73 at the mirror assembly interface 74.
Wavelength division multiplexing can be accomplished in various ways. In one form, each laser in a given laser array can be of the same wavelength to facilitate the manufacturer of such arrays, but the lasers in different arrays would be of different wavelengths.
Alternatively, a single laser array can include many lasers operating at different wavelength. At the present state of the art this would complicate the manufacture of such an array. However, it is likely that advances in the state of the art will some day make such an array readily available.
In some instances, it may be desirable to have a laser array that includes a plurality of groups of lasers, all the lasers in the group emitting the same wavelength, but different groups emitting different wavelengths.
In a wavelength division multiplexing system, a number of light beams of different wavelengths would share a common light path for part of their travel through the network but ultimately they would be separated to target different photoreceivers.
To route the beams of different wavelengths on a common light path appropriately, the optical elements used to deflect the beams should be made wavelength selective as would be apparent to a worker in the art. In particular, the repeaters shown in FIGS. 3, 4 and 5 can be made wavelength selective by choice of the reflecting facets and the wavelength of the regenerating laser.
In a monochromatic system in which all the light beams are of the same wavelength, a large number of beam paths is needed. The number of beam paths needed can be reduced by using wavelength division multiplexing in which each optical path accommodates a plurality of beams each of a different wavelength.
FIG. 6 illustrates one possible arrangement 80 for coupling into an out of a free space optical network that uses three wavelengths. Three diode laser arrays 81A, 81B and 81C each of a different wavelength are coupled to the optical path by way of a separate mirror assembly 82A, 82B and 82C basically of the kind previously described except that at the interface between the two prisms of the assembly the reflecting facets are replaced by a coating 84 that acts as a narrow band reflection interference filter that deflects the intended wavelength by 90 degrees and transmits all other wavelengths. In this way the beams emitted by the three laser arrays are combined and then deflected by the mirror assembly 85 which does include at the prism interface facets 87 for deflecting the various beams vertically into the main optical channels appropriately.
Conversely, optical beams of the three different wavelengths traveling vertically in the optical channels are deflected horizontally by such reflecting facets for travel horizontally for passage through mirror assemblies 86A, 86B and 86C, which resemble mirror assemblies 82A, 82B and 82C in that each is adapted to deflect selectively different ones of the beams of different wavelengths being used but to transmit all others. In this way beams of particular wavelength may be directed at the appropriate photodetectors 89 in the manner previously described.
These embodiments similarly may be modified to provide for regeneration of multiple beams as previously described.
It should be evident that various other modifications may be made either in the optical interconnection network or the components which have been described without departing from the spirit and proper scope of the invention.
In particular, various design considerations will determine the optimum position of the repeaters in the network, the number of beams to be amplified at a particular repeater and the amount of amplification provided at a given repeater.
Schenfeld, Eugen, Redmond, Ian R.
| Patent | Priority | Assignee | Title |
| 7106973, | Aug 13 2002 | SIGNIFY HOLDING B V | Apparatus and method for use in free-space optical communication comprising optically aligned components integrated on circuit boards |
| 7416352, | Sep 08 2005 | Northrop Grumman Systems Corporation | Optical multi-channel free space interconnect |
| 8244134, | Jun 19 2007 | Hewlett Packard Enterprise Development LP | Optical interconnect |
| Patent | Priority | Assignee | Title |
| 5268787, | Feb 17 1993 | Energy Compression Research Corp. | Multiple-pass method and apparatus for laser amplification |
| 5349593, | Jul 06 1992 | Mentor Graphics Corporation | Optical regenerator using bistable fabry-perot interferometer |
| 5585967, | Sep 07 1993 | DISNEY ENTERPRISES, INC | Three dimensional virtual image system |
| 5587831, | Dec 16 1992 | BRUKER OPTICS, INC | Semiconductor foil beam-splitter for an infrared spectrometer |
| Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
| Nov 16 1995 | NEC Research Institute, Inc. | (assignment on the face of the patent) | / | |||
| Jan 21 1998 | NEC Research Institute, Inc | NEC Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 008920 | /0865 |
| Date | Maintenance Fee Events |
| Jan 08 1999 | ASPN: Payor Number Assigned. |
| Feb 08 2001 | M183: Payment of Maintenance Fee, 4th Year, Large Entity. |
| Feb 01 2005 | M1552: Payment of Maintenance Fee, 8th Year, Large Entity. |
| Jan 28 2009 | M1553: Payment of Maintenance Fee, 12th Year, Large Entity. |
| Date | Maintenance Schedule |
| Aug 26 2000 | 4 years fee payment window open |
| Feb 26 2001 | 6 months grace period start (w surcharge) |
| Aug 26 2001 | patent expiry (for year 4) |
| Aug 26 2003 | 2 years to revive unintentionally abandoned end. (for year 4) |
| Aug 26 2004 | 8 years fee payment window open |
| Feb 26 2005 | 6 months grace period start (w surcharge) |
| Aug 26 2005 | patent expiry (for year 8) |
| Aug 26 2007 | 2 years to revive unintentionally abandoned end. (for year 8) |
| Aug 26 2008 | 12 years fee payment window open |
| Feb 26 2009 | 6 months grace period start (w surcharge) |
| Aug 26 2009 | patent expiry (for year 12) |
| Aug 26 2011 | 2 years to revive unintentionally abandoned end. (for year 12) |